As the AI World Turns — Anthropic, Figma and the Great Betrayal

And now welcome to As the AI World Turns. " If you follow the AI industry as close as we follow it here at Techstrong, it starts looking like a frigging soap opera. I don't know how many of you watch soap operas.

I don't know how many of you, maybe your mom watched soap operas. " Or how many of you are old enough to remember even "As the World Turns," right? And that's what this little promo was.

But what made for the good soap operas were the characters in it. They were so extreme that they were almost unbelievable, borderline unbelievable. And the same thing goes on with our AI, right?

The people in this AI industry, and the egos, and folks we cover every day. I'm talking about people like slippery Sam Altman and Dario, "I'll look out for you," having my baby Elon Musk. Even the administration and everyone else, and of course, there's the godfather in his leather jacket, Jensen Huang.

No one would believe this. If we went to the TV stations and said, "Hey, I want to pitch a show to you about the AI industry," like "Silicon Valley" was on HBO, they would say, "Nah, this is too unbelievable. " And they wouldn't believe what's going on here.

" I call it As the AI World Turns, and this is the first installment or the first episode of this series. " But for this week's segment, we're going to focus in on our friend Dario Amodei. He's going to be the star of this one because you got to hand it to this guy.

He and Anthropic have spent unbelievably amount of money and capital in trying to prove that they're the good guys in AI. They're the AI with a conscience. They're the AI that's not going to let the government push them around and use it for illegal means or immoral means.

They're the poor guys that the government stopped with the letter. But what we've got is it's not all fun and games, and people in Silicon Valley we're reading are getting really upset with Dario. And let's look at Dario, right?

Where did Dario come from? Well, he got mad at slippery Sam over at OpenAI and decided he was going to do it better. He was going to be more principled.

He was a sober scientist warning us about all the evils of OpenAI, and he was going to do it right. Again, just another character in our soap opera. Got Elon Musk, the rival suitor about it.

And that's not even talking, by the way, about Zuck, and I mentioned Jensen before, and of course, there's Satya, everybody's good uncle who, crazy like a fox, more like the scorpion and the frog. But in any event, people in Silicon Valley are getting a little wise, it seems, to Dario Amodei and his not whining, but professing to be the good guy and how they really operate there. I want to bring out a particular incident today, which was around what happened with one of their partners.

My teleprompter's just run way the hell out, and I don't know where I am, so we should bring it really back all the way near the beginning. So what we've got, though, is people are making claims now against Anthropic that what they're doing is they're partnering up with people, and then they're seeing what you do for a business and then copying it. The perfect example is Figma.

Figma is a partner of Anthropic, and they kind of opened the kimono a bit and showed them what they had. And then lo and behold, just a couple of weeks ago, Anthropic comes out with something called Anthropic Design. And what is Anthropic Design?

Well, it's a poor man's Figma. Figma got upset, rightfully so, because who was Figma's partner for AI? Anthropic.

So think about it. Anthropic was empowering them with Anthropic AI, saw what Figma was doing, and decided to do it for himself. There was a guy, an Anthropic executive, even on the board of Figma.

By the way, he resigned three days before Anthropic announced Claude Design. Is that a guilty conscience or just rats leaving the sinking ship? " Now, let me put the legalese in here so we don't get sued.

There's no public evidence that Anthropic stole private IP of Figma or violated any contract. They certainly didn't need Figma to tell them how to build a great visual design product using AI. But this whole ecosystem, and that's what we see in this soap opera, is built largely on handshakes, on personalities, on who it is.

And trust matters. And trust is something that's at a minimum right now, I think, in the AI industry. Who in the AI industry do you trust?

Do you trust Elon? Do you trust slippery Sam? Do you trust Dario?

Do you trust the Chinese and their open-weight models? Do you trust Tim Cook in Apple's walled garden? Do you trust Satya and what Microsoft's done and given their history over the years?

Who do you trust? Do you trust Larry Ellison, right? Also, not a fan favorite over the years.

There's not a good guy in this series. There's not a good guy in the room here. So the moral of the story is today's strategic partner can quickly become tomorrow's unpaid product research department.

Right? Now you have to ask yourself, though, let's take it back. Why would Anthropic do this?

Why burn a relationship? Why burn a partner? And the answer to that lies in the indispensability trap.

I think I mentioned it on one or two before. So ubiquitous also becomes cheap, and the profit is capped. And so the people who want to make real money move up stack.

So for Anthropic, having a great model for AI is great, and the more people who use it, great. But they want to build on top of that model like Figma has, right? And build a great product on top of it.

Figma owns the interface. Figma owns the workflow. Figma owns the customer relationship.

But Anthropic is the strategic supplier. And you know what? Suppliers can be squeezed.

Products are commoditized. Their customers can move to-- And that's one of the things with harnesses and everything. Customers can skip models today like you're changing socks.

It takes the leverage away from the commodity player, and in this case, it's models. When there's Chinese open-weight models that are near or good enough compared to OpenAI and Anthropic, you're going to see Anthropic and OpenAI moving up stack, trying to desperately climb out of the commodity basement. Otherwise, they're vulnerable to being controlled.

They need to be the layers above. So Anthropic's desperately trying to escape the indispensability check. So is OpenAI, for that matter.

So is Microsoft, so is Google, so is AWS. But we as end users, right, as consumers, we sit on top of all those. And it's our choices on what models we want to use, what commodities we want to use, what integrations, MCPs we want to use.

Because these guys, they don't necessarily play by the rules, and I think that's what we're finding out in AI. And I don't mean illegal, going to jail rules. I mean sort of the norms of business rules.

Because for them, every integration offers a view into a customer behavior. Every joint product effort teaches the model provider more about the workflow. The more they run, the more they do with every successful partner, it reveals another application that that model maker may eventually decide to absorb and go into.

As I said, Anthropic's not alone in doing this. OpenAI, they're not doing just foundation models either. Go look.

They've got OpenAI Health and OpenAI all kinds of different ones, shopping, productivity, enterprise apps. Again, this is the tangled web of this soap opera. Microsoft owns a good chunk of OpenAI.

" They're developing their own models for their copilots, their own agent. Another example, Amazon. Amazon was one of the first big backers of Anthropic to the tune of like $8 million.

They've put in, I think, $25 million above that. But they're building their own alternatives. They're partnering with OpenAI and everyone else to all plug into Bedrock as the harness.

Here's one company that should have it all, Google. Google owns most of its own stack. It doesn't have to rely on anyone.

But you still see them doing deals. They're doing deals with Space AI and everyone else. Nvidia.

So Nvidia is, Jensen, as I said, is the godfather in this one. But they're expanding out of just pure chips because he realizes the scarcity of GPUs is great right now, and it's powering a $5 trillion company. But they got to move beyond GPUs.

They got to move beyond the AI factories. They got to find what's next, because to keep that kind of valuation and to keep that kind of growth, it's like heroin. You got to find what's my new drug?

So you've got chip makers moving into systems, systems companies moving into clouds, clouds moving into models, model companies moving into applications, application companies fighting to control workflows, proprietary data, and their customer relationships. Nobody wants to stay in their zone. It's like a game of musical chairs.

No one wants to get squeezed or commoditized or replaced. And that's why everybody in this industry is someone else's partner. At the same time, they're someone else's supplier, customer, investor, and existential threat.

And that's what makes it a soap opera, because they're all five at once. The other complaints I was reading in an article about Anthropic, they fit that same pattern that I mentioned here. They don't join into open source or open weight AI.

Everyone signed, not Dario. Dario says he's never advanced banning all open weight models, though I'm sure he'd like to. But he just says, less capable open models are good for the public good if you're looking to do low-level work.

But highly capable models like Claude, that's the shizzle. That's where they've got to go, especially if you want to do real work like biological, cyber, or national security capabilities. If it's meaningful, you need Anthropic.

Well, that's what Dario says. That's a little bit more nuanced than Anthropic's critics sometimes acknowledge, but it's dead on. It's also remarkably convenient for Anthropic, right?

Because they are advocating for restrictions on these Chinese open weight models, and they cite a lot of things, the Chinese menace, security risk, protecting American AI dominance, wrapping themselves in the flag. Guys, there's nothing these characters in this soap opera won't do to further their own benefit, right? I'm not saying Ahmadi's a bad person.

I'm not saying he's inventing these dangers. All I am saying is, sometimes you got to sit back, you got to laugh, and realize that this really is a soap opera. Safety can be a sincere conviction, but it's also a formidable competitive moat, right?

Another argument that Anthropic makes, every time the Chinese models come out with a new thing, he gets up there and says, "They stole from us. " Distillation. First it was DeepSeek, then Moonshot AI, now Minimax and Kimi and the rest.

And they chart out proof. But does anybody find it ironic that here's Ahmadi b******g and moaning about the Chinese distilling information from Claude when Claude itself was built on all our information? I'm a publisher here with a couple of hundred thousand articles on really good tech stuff.

You don't think that they took that without my permission, without compensating me, to build their model? No, of course they did. You don't see me getting up and yelling and screaming about it.

I'd like to get paid for my IP, but I'm not Dario Ahmadi. I guess it's okay for them to suck someone else's brain, but if their brain gets sucked, they yell and scream like little children. Here's another thing about Anthropic, the good guys.

They recently changed their licensing. They're allowed to retain your data for 30 days for their most advanced models. Now they're saying they're using it to identify jailbreaks, improve security.

And maybe that is what they're using it for, but do you believe them that that's all they're using it for? I don't. They need more fresh data to train on.

They're asking enterprises to surrender a degree of privacy and sovereignty to make Anthropic more safe. Pay me for it, Anthropic, because you're already collecting money from me. You shouldn't be getting my data, too.

Satya said that, but Satya, of course, had his own motives, because that's how it works in this soap opera. Concentrated power does not become harmless merely because the people exercising it have good intentions, and Anthropic is no better and no worse than every other player in this soap opera. I give it credit.

It showed genuine backbone when it stood up to the Pentagon, and it paid the price. Because biological weapons and cyber attacks and uncontrolled frontier capabilities are not something, frankly, I want to put in the hands of the US Defense Department or Department of War or whatever the hell they call it today. But that doesn't make Anthropic an angel, and it doesn't mean they're not cynical about what they're doing.

They're just another character in this soap opera that we're all watching play out, except this isn't just something you shut off on your TV and can't wait to see what tomorrow's episode brings. This soap opera is going to maybe affect your life, my life, our children's lives, and every other life that's going on here. So though Anthropic started as the principal challenger to OpenAI's successes, I think in our first episode of "As the AI World Turns," we're finding out they're no angels.

But there are no angels in this soap opera. It's not black and white. It's not even color.

It's all gray, different shades of gray. Anthropic is the perfect subject for this first installment. But "As the AI World Turns" doesn't exist just because the industry's roles and mores have been settled yet.

None of us know who will ultimately become the platform, the supplier, the model, the killer app, or the casualty. You have to watch every episode to see how it unfolds. All of these guys and all their companies are racing to become indispensable before the market hardens around them and freezes their potential in.

It plays itself out in the financing, too. It's all circular that would make a fiber optic salesman in the early 2000s blush. Yesterday's rebels become tomorrow's incumbent, while today's customers discovered that the most important supplier it has been developing is a remarkably familiar product.

" So stay tuned for our next episode of this. You should follow along because you couldn't make this stuff up. But for now, that's what I think.

I'm Shimmy, I'm out. We'll see you next week. Shimmy says, Shimmy says, Shimmy, Shimmy, Shimmy says, ask me almost anything!